Commits · 770a22e997ad70a949a317784c027d66a35a4626 · KMSCAKKSCFKA AKFACAMADCAS / tensorflow-federated

This project is mirrored from https://github.com/tensorflow/federated. Pull mirroring failed 1 year ago.
Repository mirroring has been paused due to too many failed attempts. It can be resumed by a project maintainer or owner.
Last successful update 1 year ago.

Jul 21, 2020

Update FLARS code to work with . · 770a22e9

Zachary Garrett authored 4 years ago

Also while were here, update to some more modern TFF coding practices:
-  Use `tff.learning.framework.ModelWeights` instead of a custom type.
-  Delete `from_tff_result` conversion helper, it is no longer needed.
-  Remove unused test helper methods.
-  Replace numpy dependency with TF ops.

PiperOrigin-RevId: 322359152

770a22e9

Jul 15, 2020
- Fix typos by replacing `TrainableModel` `Model`. · 4fd7f46f
  Zachary Garrett authored 4 years ago
  
  TrainableModel has been removed from TFF for a while. PiperOrigin-RevId: 321184501
  4fd7f46f
Jun 02, 2020

Change Keras metric aggregation to produce a single federated aggregate over · f022e33f

Zachary Garrett authored 4 years ago

the "zipped" metrics, rather than zipping the results of many federated
aggregates.

This has been seen to significantly speed up execution, as much as 2x for 50
clients and 10 metrics.

Thanks to @jkr26 for investigation!

PiperOrigin-RevId: 314207362

f022e33f

May 28, 2020
- Remove python lint as directives, part 2. · 9450611d
  Michael Reneer authored 4 years ago
  
  PiperOrigin-RevId: 313419009
  9450611d
Apr 14, 2020
- Move usage of `tff.utils.IterativeProcess` to `tff.templates.IterativeProcess`. · d7c8e3a2
  Michael Reneer authored 4 years ago
  
  PiperOrigin-RevId: 306446003
  d7c8e3a2
Feb 19, 2020

Remove calls to deprecated tff.learning.from_compiled_keras_model(). · 8a8e1e45

Zachary Garrett authored 5 years ago

-   Removes calls to `tff.learning.from_compiled_keras_model()`.
-   Change ModelWeights to use lists rather than a dictionaries, remove flattening
    of variables and gradients in training loop.
-   Add parameter for client optimizers to federated computation building.
-   Update to use kwargs constructor for OrderdDict.
-   Update tf.keras.Sequential model construction to explicit use an Input
    layer instead of using an input_shape parameter.
-   Rplace the check for TrainableModel with a regular Model check in
    DummyClientComputationFn, since the optimizer is no longer compiled in.

PiperOrigin-RevId: 295772881

8a8e1e45

Feb 14, 2020

Refactoring after bumping `attr` package version. · 02d783f6

Zachary Garrett authored 5 years ago

Migrates `cmp` arguments in attr.s() to `eq` and `order` arguments:
1. If `cmp=False`, replacing it with `eq=False`.
2. If `cmp=True`, replacing it with `eq=True, order=True`.

PiperOrigin-RevId: 294921678

02d783f6

Dec 21, 2019
- Rename from_anon_tuple to from_tff_result. · e60b4463
  Brendan McMahan authored 5 years ago
  
  PiperOrigin-RevId: 286616288
  e60b4463
Nov 21, 2019
- Replace calls to tff.federated_apply with tff.federated_map. · b1c2ab5a
  Zachary Charles authored 5 years ago
  
  PiperOrigin-RevId: 281621220
  b1c2ab5a
Nov 16, 2019
- Automated rollback of commit 22fc942d · ce74020e
  A. Unique TensorFlower authored 5 years ago
  
  PiperOrigin-RevId: 280784907
  ce74020e
- Internal · 22fc942d
  Michael Reneer authored 5 years ago
  
  PiperOrigin-RevId: 280767713
  22fc942d
Nov 15, 2019
- Fix error in FLARS experiments. · 4216c090
  Zachary Garrett authored 5 years ago
  
  PiperOrigin-RevId: 280651562
  4216c090
Nov 09, 2019
- Open source initial FLARS research. · 31ce02ff
  Zachary Garrett authored 5 years ago
  
  Implementation by @slowbull during time at Google. PiperOrigin-RevId: 279398722
  31ce02ff
Nov 07, 2019
- Flatten out simple_fedavg further, removing internal dependencies. · 8c904235
  Zachary Garrett authored 5 years ago
  
  PiperOrigin-RevId: 278962151
  8c904235
Oct 04, 2019
- Used tf.Variable.assign() instead of tf.assign() to be TF 2.0 compliant. · 1b69104a
  Zachary Garrett authored 5 years ago
  
  PiperOrigin-RevId: 272689129
  1b69104a
Sep 13, 2019
- Move `baseline_fedavg` package to `simple_fedavg`. · 5a93388b
  Michael Reneer authored 5 years ago
  
  PiperOrigin-RevId: 268790969
  5a93388b
Sep 12, 2019

Update the `research` package to use the pip-style import. · fceaea71

Michael Reneer authored 5 years ago

* Updated the `research` package imports.
* Increased the BUILD visibility strictness.

PiperOrigin-RevId: 268692104

fceaea71

Jul 08, 2019
- Simplifies implementation of simple_fedavg research baseline and cleans up documentation. · 000e9aa6
  Jakub Konecny authored 5 years ago
  
  PiperOrigin-RevId: 256903837
  000e9aa6
Jun 28, 2019

Add new FedAvg implementation for research. · 0f0d7872

Zhouyuan Huo authored 5 years ago

To modify the code, we can modify how client computes in "client_update", and modify how server computes in "server_update".

PiperOrigin-RevId: 255438558

0f0d7872